Re: web+ and registerProtocolHandler

On Wed, Sep 19, 2012 at 11:51 AM, Mark Nottingham <mnot@mnot.net> wrote:
> On 19/09/2012, at 1:10 AM, Robin Berjon <robin@berjon.com> wrote:
>> On Sep 18, 2012, at 22:39 , Mark Nottingham <mnot@mnot.net> wrote:
>>> If I were trying to solve this problem, I'd be allowing people to register handlers for link *relations*, not schemes; has that come up at all? After all, OpenID is already coordinated through relations…
>>
>> Maybe I'm missing something, but this is intended to be system-wide. How do you convey a link relation in email for instance? Or over Twitter?
>
> Understood, but what's the use case?
>
> I'm not saying that registerProtocolHandler isn't necessary; if somebody wants to send a telnet: link in Twitter and have their favourite Web telnet client open, that's cool.
>
> What I am saying is that using the URI scheme to indicate *any* new protocol -- when by "protocol" you mean a coordination of HTTP, like OAuth or OpenID -- isn't what URI schemes are really for, and I have concerns about where this kind of design will take us (when anybody coming up with a new kind of "web service" decides to mint a URI scheme to identify it).
>
> That sort of thing *is* much more natural to express as a link relation; has anyone proposed registerRelationHandler()?
>
> As to expressing *that* sort of thing in a bare (i.e., context-free) link -- is there really a use case? I.e., how useful is it to express a bare OpenID endpoint in a twitter link, vs. a link to a Web page that then provides an entry point to an OpenID interaction?

As I wrote before, the folks who are best able to answer these
questions are not reading this thread.

Adam